基于多知识的Web网页信息抽取方法

INFORMATION EXTRACTION FROM Web PAGES BASED ON MULTI KNOWLEDGE

从 Web网页中自动抽取所需要的信息内容 ,是互联网信息智能搜取的一个重要研究课题 .为有效解决网页信息抽取所需的信息描述知识获取问题 ,这里提出了一种基于多知识的 Web网页信息抽取方法 (简称 MKIE方法 ) .该方法将网页信息抽取所需的知识分为二类 .一类是描绘网页内容本身表示特点 ,以及识别各网页信息对象的确定模式知识 ;另一类则描述网页信息记录块 ,以及各网页信息对象的非确定模式知识 .MKIE方法根据前一类知识 ,动态分析获得后一类知识 ;并利用这两类知识 ,最终完成从信息内容类似但其表现形式各异的网页中 ,抽取出所需要的信息 .美大学教员论文网页信息抽取实验结果表明 。

Web page information extraction is an important study field in the AI research on the WWW. To tackle the problem of knowledge acquisition in the web page information extraction, a new approach based on multi knowledge for web information extraction is put forward. Two knowledge are used in this method (called MKIE). One is composed of the knowledge that characterizes the web page content and the definite patterns for the information objects in the web pages. The other one consists of the knowledge that describes the web page content displaying patterns and the non definite patterns for the information objects in the web pages. The latter knowledge can be acquired through learning based the former type knowledge and those two knowledge are used together to extract information from the same sort of web pages with different displaying styles successfully. The experiment made on the U.S. university faculty's paper web page information extraction show that the MKIE has a powerful ability to recognize the information displaying patterns and to extract them effectively.